Merge pull request #356 from bluescarni/pr/relu

Add ReLU

CMakeLists.txt

@@ -270,6 +270,7 @@ set(HEYOKA_SRC_FILES
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/log.cpp"
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/pow.cpp"
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/sigmoid.cpp"
+    "${CMAKE_CURRENT_SOURCE_DIR}/src/math/relu.cpp"
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/sin.cpp"
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/sqrt.cpp"
     "${CMAKE_CURRENT_SOURCE_DIR}/src/math/tan.cpp"

doc/changelog.rst

@@ -7,6 +7,8 @@ Changelog
 New
 ~~~
 
+- Implement ``ReLU`` and its derivative in the expression
+  system (`#356 <https://github.com/bluescarni/heyoka/pull/356>`__).
 - Implement the eccentric longitude :math:`F` in the expression
   system (`#352 <https://github.com/bluescarni/heyoka/pull/352>`__).
 - Implement the delta eccentric anomaly :math:`\Delta E` in the expression
@@ -24,6 +26,8 @@ Changes
 Fix
 ~~~
 
+- Fix compiler warning when building without SLEEF support
+  (`#356 <https://github.com/bluescarni/heyoka/pull/356>`__).
 - Improve the numerical stability of the VSOP2013 model
   (`#353 <https://github.com/bluescarni/heyoka/pull/353>`__).
 - Improve the numerical stability of the Kepler solver

include/heyoka/math.hpp

@@ -27,6 +27,7 @@
 #include <heyoka/math/log.hpp>
 #include <heyoka/math/pow.hpp>
 #include <heyoka/math/prod.hpp>
+#include <heyoka/math/relu.hpp>
 #include <heyoka/math/sigmoid.hpp>
 #include <heyoka/math/sin.hpp>
 #include <heyoka/math/sinh.hpp>

include/heyoka/math/relu.hpp

@@ -0,0 +1,97 @@
+// Copyright 2020, 2021, 2022, 2023 Francesco Biscani ([email protected]), Dario Izzo ([email protected])
+//
+// This file is part of the heyoka library.
+//
+// This Source Code Form is subject to the terms of the Mozilla
+// Public License v. 2.0. If a copy of the MPL was not distributed
+// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+#ifndef HEYOKA_MATH_RELU_HPP
+#define HEYOKA_MATH_RELU_HPP
+
+#include <cstdint>
+#include <vector>
+
+#include <heyoka/config.hpp>
+#include <heyoka/detail/fwd_decl.hpp>
+#include <heyoka/detail/llvm_fwd.hpp>
+#include <heyoka/detail/visibility.hpp>
+#include <heyoka/func.hpp>
+#include <heyoka/s11n.hpp>
+
+HEYOKA_BEGIN_NAMESPACE
+
+namespace detail
+{
+
+class HEYOKA_DLL_PUBLIC relu_impl : public func_base
+{
+    friend class boost::serialization::access;
+    template <typename Archive>
+    void serialize(Archive &ar, unsigned)
+    {
+        ar &boost::serialization::base_object<func_base>(*this);
+    }
+
+public:
+    relu_impl();
+    explicit relu_impl(expression);
+
+    [[nodiscard]] expression normalise() const;
+
+    [[nodiscard]] std::vector<expression> gradient() const;
+
+    [[nodiscard]] llvm::Value *llvm_eval(llvm_state &, llvm::Type *, const std::vector<llvm::Value *> &, llvm::Value *,
+                                         llvm::Value *, llvm::Value *, std::uint32_t, bool) const;
+
+    [[nodiscard]] llvm::Function *llvm_c_eval_func(llvm_state &, llvm::Type *, std::uint32_t, bool) const;
+
+    llvm::Value *taylor_diff(llvm_state &, llvm::Type *, const std::vector<std::uint32_t> &,
+                             const std::vector<llvm::Value *> &, llvm::Value *, llvm::Value *, std::uint32_t,
+                             std::uint32_t, std::uint32_t, std::uint32_t, bool) const;
+
+    llvm::Function *taylor_c_diff_func(llvm_state &, llvm::Type *, std::uint32_t, std::uint32_t, bool) const;
+};
+
+class HEYOKA_DLL_PUBLIC relup_impl : public func_base
+{
+    friend class boost::serialization::access;
+    template <typename Archive>
+    void serialize(Archive &ar, unsigned)
+    {
+        ar &boost::serialization::base_object<func_base>(*this);
+    }
+
+public:
+    relup_impl();
+    explicit relup_impl(expression);
+
+    [[nodiscard]] expression normalise() const;
+
+    [[nodiscard]] std::vector<expression> gradient() const;
+
+    [[nodiscard]] llvm::Value *llvm_eval(llvm_state &, llvm::Type *, const std::vector<llvm::Value *> &, llvm::Value *,
+                                         llvm::Value *, llvm::Value *, std::uint32_t, bool) const;
+
+    [[nodiscard]] llvm::Function *llvm_c_eval_func(llvm_state &, llvm::Type *, std::uint32_t, bool) const;
+
+    llvm::Value *taylor_diff(llvm_state &, llvm::Type *, const std::vector<std::uint32_t> &,
+                             const std::vector<llvm::Value *> &, llvm::Value *, llvm::Value *, std::uint32_t,
+                             std::uint32_t, std::uint32_t, std::uint32_t, bool) const;
+
+    llvm::Function *taylor_c_diff_func(llvm_state &, llvm::Type *, std::uint32_t, std::uint32_t, bool) const;
+};
+
+} // namespace detail
+
+HEYOKA_DLL_PUBLIC expression relu(expression);
+
+HEYOKA_DLL_PUBLIC expression relup(expression);
+
+HEYOKA_END_NAMESPACE
+
+HEYOKA_S11N_FUNC_EXPORT_KEY(heyoka::detail::relu_impl)
+
+HEYOKA_S11N_FUNC_EXPORT_KEY(heyoka::detail::relup_impl)
+
+#endif

include/heyoka/math/sin.hpp

@@ -9,7 +9,7 @@
 #ifndef HEYOKA_MATH_SIN_HPP
 #define HEYOKA_MATH_SIN_HPP
 
-#include <cstddef>
+#include <cstdint>
 #include <string>
 #include <unordered_map>
 #include <vector>

src/detail/vector_math.cpp

@@ -31,6 +31,10 @@ namespace
 
 using vf_map_t = std::unordered_map<std::string, std::vector<vf_info>>;
 
+// NOTE: make_vfinfo() is necessary if *any* vectorisation backend is active,
+// but at the moment we have only SLEEF.
+#if defined(HEYOKA_WITH_SLEEF)
+
 auto make_vfinfo(const char *s_name, std::string v_name, std::uint32_t width, std::uint32_t nargs)
 {
     assert(nargs == 1u || nargs == 2u);
@@ -40,6 +44,8 @@ auto make_vfinfo(const char *s_name, std::string v_name, std::uint32_t width, st
     return ret;
 }
 
+#endif
+
 #if defined(HEYOKA_WITH_SLEEF)
 
 // NOLINTNEXTLINE(bugprone-easily-swappable-parameters)

src/math/cos.cpp

@@ -22,7 +22,7 @@
 
 #include <boost/numeric/conversion/cast.hpp>
 
-#include <fmt/format.h>
+#include <fmt/core.h>
 
 #include <llvm/IR/BasicBlock.h>
 #include <llvm/IR/DerivedTypes.h>
@@ -223,6 +223,8 @@ llvm::Value *taylor_diff_cos_impl(llvm_state &s, llvm::Type *fp_t, const cos_imp
     return llvm_fdiv(s, ret_acc, div);
 }
 
+// LCOV_EXCL_START
+
 // All the other cases.
 template <typename U, std::enable_if_t<!is_num_param_v<U>, int> = 0>
 llvm::Value *taylor_diff_cos_impl(llvm_state &, llvm::Type *, const cos_impl &, const std::vector<std::uint32_t> &,
@@ -233,6 +235,8 @@ llvm::Value *taylor_diff_cos_impl(llvm_state &, llvm::Type *, const cos_impl &,
         "An invalid argument type was encountered while trying to build the Taylor derivative of a cosine");
 }
 
+// LCOV_EXCL_STOP
+
 llvm::Value *taylor_diff_cos(llvm_state &s, llvm::Type *fp_t, const cos_impl &f, const std::vector<std::uint32_t> &deps,
                              const std::vector<llvm::Value *> &arr, llvm::Value *par_ptr, std::uint32_t n_uvars,
                              std::uint32_t order, std::uint32_t idx, std::uint32_t batch_size)
@@ -370,6 +374,8 @@ llvm::Function *taylor_c_diff_func_cos_impl(llvm_state &s, llvm::Type *fp_t, con
     return f;
 }
 
+// LCOV_EXCL_START
+
 // All the other cases.
 template <typename U, std::enable_if_t<!is_num_param_v<U>, int> = 0>
 llvm::Function *taylor_c_diff_func_cos_impl(llvm_state &, llvm::Type *, const cos_impl &, const U &, std::uint32_t,
@@ -379,6 +385,8 @@ llvm::Function *taylor_c_diff_func_cos_impl(llvm_state &, llvm::Type *, const co
                                 "of a cosine in compact mode");
 }
 
+// LCOV_EXCL_STOP
+
 llvm::Function *taylor_c_diff_func_cos(llvm_state &s, llvm::Type *fp_t, const cos_impl &fn, std::uint32_t n_uvars,
                                        std::uint32_t batch_size)
 {